Workflows, pipelines and the order to build them in

The documentation defines the two pieces everything is built from:

“Workflow automation is the process of using technology to perform repetitive tasks automatically.” “A trigger is an event that sets your workflow into motion.” “An action is the task performed in response to a trigger.”

The documented examples are deliberately ordinary — a form submission, a booked appointment, a reply to a message, then an email sent or a lead assigned. That is the right expectation to start from. The value is not in one clever automation; it is in removing the handful of steps a person currently performs from memory, several times a day, across several clients.

Automation acts on records; pipelines are where you see them move:

“Pipelines are visual workflow tools that show the progress of leads, deals, or customers from start to finish.” “A pipeline is made up of stages, with each stage representing a step in your sales or customer journey.”

For an agency there are always at least two pipelines and they are easy to confuse. One is your own new business — prospect to proposal to signed. The other belongs to each client and tracks their leads. Same mechanics, different owner, and mixing them is the most common setup mistake because in the first week both are small enough to look like one list.

1. Name stages after what happened, not how you feel“Proposal sent” is a fact; “warm” is an opinion that means something different to each person on the team. Stages that describe events survive a handover.
2. Keep the count low enough to be filled in honestlyFive or six stages that everyone updates beats twelve that nobody does. A stage exists so an automation can fire and a report can be true — neither works if the pipeline is half-maintained.
3. Decide what moves a record, and whether a person does itIf a stage change is meant to be automatic, build the trigger. If it is manual, say who does it and when. Stages nobody owns are the reason dashboards stop matching reality by month three.

1. First: the follow-up that depends on somebody rememberingA new enquiry that gets a reply within minutes rather than when someone next checks. This is the automation that pays for itself immediately, and the one most agencies build fifth.
2. Second: appointment confirmation and remindersLow effort, immediately visible to the client, and the easiest thing to point at as evidence that the new system is doing something. It also reduces the no-shows that make a calendar look busier than the week actually was.
3. Third: the handover between two peopleWhatever currently travels as a message — “can you pick this up” — becomes an assignment with a record. You are not saving keystrokes; you are removing the step where things get dropped.
4. Fourth: client onboardingThe sequence every new client gets: the welcome, the information request, the kickoff booking, the chase when the information does not arrive. This is the one worth templating, because you will run it again for the next client.
5. Only then: reporting and the clever onesThey take longest to build and break most often. Build them once the first four have run for a month without anyone touching them.

Break it on purposeOnce a workflow works, make it fail deliberately — a bad email address, a missing field — and see whether anyone finds out. An automation that fails silently is worse than the manual step it replaced, because the manual step had a person attached who noticed.

The economics of agency automation only work if the build is reusable. Otherwise every new client re-runs the same fortnight, and the platform has moved the work rather than removed it.

The documented mechanism is the snapshot, and its exclusions matter as much as its contents:

“Snapshots are reusable templates that capture selected configuration assets from a HighLevel sub-account so they can be copied into new or existing sub-accounts.” They carry workflows, triggers, calendars, custom fields, pipelines, tags, forms, funnels and dashboards. They are “designed to transfer configuration and selected reusable assets, not live account activity or customer records”: contacts, appointments, conversations and message history, reputation data, Stripe connections, integrations and assigned phone numbers do not transfer.

So the template carries your workflows, pipelines, calendars, forms and custom fields — the whole shape of a client setup. It does not carry contacts, conversations, appointments, phone numbers or integrations. Onboarding is therefore always two jobs: apply the configuration, then bring the data and reconnect the accounts. Budget for the second one; it is where the fortnight actually goes.

Version your templateA snapshot is a copy taken at a moment, not a live link. Improving a workflow in one client’s sub-account does not update the others or the template. Decide where the canonical version lives before you have eight copies drifting apart.

It will not fix a process nobody has agreed on. Automating a disputed handover just makes the dispute run faster and more often. Nor will it fix data: a workflow triggered by a field half your team fills in inconsistently inherits that inconsistency and applies it at scale, across every client at once.

Both are cheap to settle before you build and expensive afterwards, because by then the broken version is running in several sub-accounts and each one needs correcting separately.